-
Notifications
You must be signed in to change notification settings - Fork 1
/
2d_param_8fold_TvsAN.R
263 lines (227 loc) · 12.3 KB
/
2d_param_8fold_TvsAN.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
#################################################
# prepare workspace
#################################################
args <- commandArgs(trailingOnly = TRUE)
beg <- as.numeric(args[1]) # first ID to process
end <- as.numeric(args[2]) # last consequitve ID to process
# data <- args[1]
load("./data_BRCA.RData") # data to work on, includes folds_an and folds_t
#################################################
# libs
#################################################
library(dgRaph)
library(dplyr)
library(VGAM)
library(bbmle)
library(pROC) # calculates AUC
loglik <- function(alpha, beta){
-sum(dbetabinom.ab(xi, ni, alpha, beta, log = T))
}
#IDs <- c("RAG1AP1","CPA1","NEK2","RNASEH2A","LOC148145","TMEM63B","TIMM17A","PLK1","RABIF","PTF1A")
# setup variables to hold model and performance metrics across folds
#facPotNormals <- NULL
#facPotTumours <- NULL
#models <- NULL # a list to hold models and performance measures across folds
for (i in beg:end) {#2:length(IDs)) { # iterate through given jobs
#i <- which(names(data_BRCA) %in% IDs[j]) # find the index to work with
cat(paste("doing ",i,"\n",sep=""))
ptm <- proc.time()[3]
#models[[j]] <- list()
#models[[j]]$gene_name <- IDs[j]
model <- list()
#model$gene_name <- IDs[j]
scores_xval <- vector(length=nrow(data_BRCA[[i]]))
# calculate common discretization upper and lower limits
expr <- as.data.frame(data_BRCA[[i]]) %>%
mutate(EXPR = read_count / lib_size) %>%
select(EXPR)
pr <- as.data.frame(data_BRCA[[i]]) %>%
select(starts_with("pr", ignore.case = F))
n_pr_cpg <- ncol(pr)
gb <- as.data.frame(data_BRCA[[i]]) %>%
select(starts_with("gb", ignore.case = F))
n_gb_cpg <- ncol(gb)
# upper and lower limits
min_p <- min(pr)-0.1
max_p <- max(pr)+0.1
min_gb <- min(gb)-0.1
max_gb <- max(gb)+0.1
min_e <- max(0,min(expr)-0.001*min(expr))
max_e <- max(expr)+0.001*max(expr)
# begin x-fold here
for (fold in 1:8) { # iterate through 8 folds
# identify training and validation sets
an_xfold_training <- setdiff((1:82),unlist(folds_an[[fold]]))
t_xfold_training <- setdiff(82+(1:730),unlist(folds_t[[fold]]))
#################################################
# Normals model
#################################################
data_normals_train <- as.data.frame(data_BRCA[[i]][an_xfold_training,])
#################################################
# Learn betabinomial
#################################################
ni <- as.integer(data_normals_train$lib_size)
xi <- as.integer(data_normals_train$read_count+1)
m0 <- mle2(minuslogl = loglik, start = list(alpha = 1, beta = 1), method = "L-BFGS-B", lower=c(alpha = 0.0001, beta = 0.0001))
alpha <- coef(m0)['alpha']
beta <- coef(m0)['beta']
# Posteriors
alphaPost <- alpha + data_normals_train$read_count
betaPost <- beta + data_normals_train$lib_size - data_normals_train$read_count
#################################################
# Data-massage
#################################################
df_normals <- data_normals_train %>%
mutate(EXPR = NA) %>%
mutate(PR_overall = NA) %>%
mutate(GB_overall = NA) %>%
mutate_each(funs(bin = as.integer(cut(., breaks = seq(min_p,max_p,length.out = 101), labels = c(1:100)))), PR = starts_with("pr", ignore.case = F)) %>%
mutate_each(funs(bin = as.integer(cut(., breaks = seq(min_gb,max_gb,length.out = 101), labels = c(1:100)))), GB = starts_with("gb", ignore.case = F)) %>%
select(-starts_with("pr", ignore.case = F), -starts_with("gb", ignore.case = F), -lib_size, -read_count)
#################################################
# Building models and training
#################################################
# Build Model
varDim <- rep(100, 3+n_pr_cpg+n_gb_cpg)
if (fold == 1) {
facPot <- list(linregPotential(dim = c(100, 100)), # PR_overall | EXPR
linregPotential(dim = c(100, 100)), # GB_overall | EXPR
linregPotential(dim = c(100, 100), range1 = c(min_p, max_p), range2 = c(min_p,max_p), alpha = 1, beta = 0, var = 0.14**2), # PR_i | PR
linregPotential(dim = c(100, 100), range1 = c(min_gb, max_gb), range2 = c(min_gb,max_gb), alpha = 1, beta = 0, var = 0.14**2)) # GB_i | GB
} else facPot <- prev_normals_potentials
facNbs <- c(list(c(1,2)), # PR_overall | EXPR
list(c(1,3)), # GB_overall | EXPR
lapply(4:(3+n_pr_cpg), FUN=function(i){c(2,i)}), # PR_i | PR_overall
lapply((1+3+n_pr_cpg):(3+n_pr_cpg+n_gb_cpg), FUN=function(i){c(3,i)}) # GB_i | GB_overall
)
potMap <- c(1, 2, rep(3, n_pr_cpg), rep(4, n_gb_cpg))
dfg_normals <- dfg(varDim, facPot, facNbs, potMap, varNames = names(df_normals))
optimFun <- list(linreg1 = linregOptimize(range1 = c(min_e,max_e), range2 = c(min_p,max_p)),
linreg2 = linregOptimize(range1 = c(min_e,max_e), range2 = c(min_gb,max_gb)))
# Data
dataList <- list()
dataList[[1]] <- lapply(1:nrow(df_normals), FUN=function(i){
breaks <- seq(min_e, max_e, length.out = 101)
diff( pbeta(breaks, alphaPost[i], betaPost[i]))
})
# Train
dfg_normals <- train(data = df_normals,
dataList = dataList,
dfg = dfg_normals,
optim = c("linreg1", "linreg2", "noopt", "noopt"),
optimFun = optimFun, iter.max = 5000, threshold = 1e-5)
prev_normals_potentials <- potentials(dfg_normals)
# Hack Expression prior
cur_length <- length(dfg_normals$facNbs)+1
facNbs[[cur_length]] <- 1
potMap[cur_length] <- 5
facPot <- potentials(dfg_normals)
facPot[[5]] <- betaPotential(range=c(min_e,max_e),alphas=alpha, betas=beta)
dfg_normals <- dfg(varDim, facPot, facNbs, potMap, varNames = names(df_normals))
#facPotNormals[[j]] <- list()
#facPotNormals[[j]][[fold]] <- potentials(dfg_normals)
#################################################
# Tumours model
#################################################
data_tumours_train <- as.data.frame(data_BRCA[[i]][t_xfold_training,])
#################################################
# Learn betabinomial
#################################################
ni <- as.integer(data_tumours_train$lib_size)
xi <- as.integer(data_tumours_train$read_count+1)
m0 <- mle2(minuslogl = loglik, start = list(alpha = 1, beta = 1), method = "L-BFGS-B", lower=c(alpha = 0.0001, beta = 0.0001))
alpha <- coef(m0)['alpha']
beta <- coef(m0)['beta']
# Posteriors
alphaPost <- alpha + data_tumours_train$read_count
betaPost <- beta + data_tumours_train$lib_size - data_tumours_train$read_count
#################################################
# Data-massage
#################################################
df_tumours <- data_tumours_train %>%
mutate(EXPR = NA) %>%
mutate(PR_overall = NA) %>%
mutate(GB_overall = NA) %>%
mutate_each(funs(bin = as.integer(cut(., breaks = seq(min_p,max_p,length.out = 101), labels = c(1:100)))), PR = starts_with("pr", ignore.case = F)) %>%
mutate_each(funs(bin = as.integer(cut(., breaks = seq(min_gb,max_gb,length.out = 101), labels = c(1:100)))), GB = starts_with("gb", ignore.case = F)) %>%
select(-starts_with("pr", ignore.case = F), -starts_with("gb", ignore.case = F), -lib_size, -read_count)
#################################################
# Building models and training
#################################################
# Build Model
if (fold != 1) facPot <- prev_tumours_potentials
facNbs[[cur_length]] <- NULL
potMap <- potMap[-cur_length]
facPot[[5]] <- NULL
dfg_tumours <- dfg(varDim, facPot, facNbs, potMap, varNames = names(df_tumours))
# Data
dataList <- list()
dataList[[1]] <- lapply(1:nrow(df_tumours), FUN=function(i){
breaks <- seq(min_e, max_e, length.out = 101)
diff( pbeta(breaks, alphaPost[i], betaPost[i]))
})
# Train
dfg_tumours <- train(data = df_tumours,
dataList = dataList,
dfg = dfg_tumours,
optim = c("linreg1", "linreg2", "noopt", "noopt"),
optimFun = optimFun, iter.max = 5000, threshold = 1e-5)
prev_tumours_potentials <- potentials(dfg_tumours)
# Hack Expression prior
cur_length <- length(dfg_tumours$facNbs)+1
facNbs[[cur_length]] <- 1
potMap[cur_length] <- 5
facPot <- potentials(dfg_tumours)
facPot[[5]] <- betaPotential(range=c(min_e,max_e),alphas=alpha, betas=beta)
dfg_tumours <- dfg(varDim, facPot, facNbs, potMap, varNames = names(df_tumours))
#facPotTumours[[j]] <- list()
#facPotTumours[[j]][[fold]] <- potentials(dfg_tumours)
#################################################
# Evaluation and performance
#################################################
# training data performance
data_train <- as.data.frame(data_BRCA[[i]][c(an_xfold_training,t_xfold_training),])
df_train <- data_train %>%
mutate(EXPR = (read_count+1) / lib_size) %>%
mutate(EXPR = as.integer(cut(EXPR, breaks = seq(min_e,max_e,length.out = 101), labels = c(1:100)))) %>%
mutate(PR_overall = NA) %>%
mutate(GB_overall = NA) %>%
mutate_each(funs(bin = as.integer(cut(., breaks = seq(min_p,max_p,length.out = 101), labels = c(1:100)))), PR = starts_with("pr", ignore.case = F)) %>%
mutate_each(funs(bin = as.integer(cut(., breaks = seq(min_gb,max_gb,length.out = 101), labels = c(1:100)))), GB = starts_with("gb", ignore.case = F)) %>%
select(-starts_with("pr", ignore.case = F), -starts_with("gb", ignore.case = F), -lib_size, -read_count)
likelihood1 <- likelihood(dfg = dfg_tumours, data = df_train, log = T)
likelihood1[which(is.na(likelihood1))] <- vapply(likelihood1[which(is.na(likelihood1))], FUN= function(x) rnorm(n=1,mean=-500), FUN.VALUE = rnorm(n=1,mean=-500))
likelihood2 <- likelihood(dfg = dfg_normals, data = df_train, log = T)
likelihood2[which(is.na(likelihood2))] <- vapply(likelihood2[which(is.na(likelihood2))], FUN= function(x) rnorm(n=1,mean=-500), FUN.VALUE = rnorm(n=1,mean=-500))
scores_train <- likelihood1 - likelihood2
# evaluation data performance
data_eval <- as.data.frame(data_BRCA[[i]][-c(an_xfold_training,t_xfold_training),])
df_eval <- data_eval %>%
mutate(EXPR = (read_count+1) / lib_size) %>%
mutate(EXPR = as.integer(cut(EXPR, breaks = seq(min_e,max_e,length.out = 101), labels = c(1:100)))) %>%
mutate(PR_overall = NA) %>%
mutate(GB_overall = NA) %>%
mutate_each(funs(bin = as.integer(cut(., breaks = seq(min_p,max_p,length.out = 101), labels = c(1:100)))), PR = starts_with("pr", ignore.case = F)) %>%
mutate_each(funs(bin = as.integer(cut(., breaks = seq(min_gb,max_gb,length.out = 101), labels = c(1:100)))), GB = starts_with("gb", ignore.case = F)) %>%
select(-starts_with("pr", ignore.case = F), -starts_with("gb", ignore.case = F), -lib_size, -read_count)
#scores_eval <- likelihood(dfg = dfg_tumours, data = df_eval, log = T) - likelihood(dfg = dfg_normals, data = df_eval, log = T)
likelihood1 <- likelihood(dfg = dfg_tumours, data = df_eval, log = T)
likelihood1[which(is.na(likelihood1))] <- vapply(likelihood1[which(is.na(likelihood1))], FUN= function(x) rnorm(n=1,mean=-500), FUN.VALUE = rnorm(n=1,mean=-500))
likelihood2 <- likelihood(dfg = dfg_normals, data = df_eval, log = T)
likelihood2[which(is.na(likelihood2))] <- vapply(likelihood2[which(is.na(likelihood2))], FUN= function(x) rnorm(n=1,mean=-500), FUN.VALUE = rnorm(n=1,mean=-500))
scores_eval <- likelihood1 - likelihood2
# Metrics
# models[[j]]$auc_training[fold] <- auc(predictor=scores_train,response=c(rep("N",length(an_xfold_training)),rep("T",length(t_xfold_training))))
# models[[j]]$auc_evaluation[fold] <- auc(predictor=scores_eval,response=c(rep("N",(82-length(an_xfold_training))),rep("T",(730-length(t_xfold_training)))))
# models[[j]]$kls[fold] <- kl(dfg_normals,dfg_tumours)
model$auc_training[fold] <- auc(predictor=scores_train,response=c(rep("N",length(an_xfold_training)),rep("T",length(t_xfold_training))))
model$auc_evaluation[fold] <- auc(predictor=scores_eval,response=c(rep("N",(82-length(an_xfold_training))),rep("T",(730-length(t_xfold_training)))))
model$kls[fold] <- kl(dfg_normals,dfg_tumours)
scores_xval[c(folds_an[[fold]],folds_t[[fold]])] <- scores_eval
}
# end x-fold here
#models[[j]]$scores <- scores_xval
model$scores <- scores_xval
eval(parse(text=paste('save(model, file="./',i,'_model.RData")',sep="")))
cat(paste("done evaluating for ",i," in ", sprintf("%.2f", (proc.time()[3]-ptm)/60)," minutes\n",sep=""))
}